Researchers at Mass General Brigham in Boston have developed artificial intelligence models that predict cancer patients’ biological age and chances of survival using just a facial photograph. This study, recently published in The Lancet Digital Health, could assist doctors in evaluating treatment options for cancer patients.
Dr. Raymond Mak, a radiation oncologist and lead author of the study, expressed surprise at the findings, stating, “I was pretty surprised to find out that we could pick up signs of mortality risk from just a facial photograph.” The AI models were trained on over 40 million facial images and assess various facial features to gauge health.
The two models, FAHR-FaceAge and FAHR-FaceSurvival, provide distinct assessments. FAHR-FaceAge estimates biological age based on facial features, while FAHR-FaceSurvival predicts survival likelihood. FAHR-FaceAge considers features around the temples and forehead, while FAHR-FaceSurvival focuses on the bridge of the nose and area beneath the eyes.
Among 122 cancer patients identified by the AI as high-risk, 52.5% died within 90 days, compared to only 2% from the low-risk group. A biological age calculation more than 20 years older than chronological age also correlated with a higher risk of mortality.
Dr. Hugo Aerts, another study co-author, noted, “To our knowledge, this is the first demonstration that an AI model can directly predict cancer survival from a simple photograph.” However, there are calls for further validation of the tool’s effectiveness across different populations and cancer types.
While the AI technology shows promise, experts cautioned that it is not intended to replace clinical evaluations. Dr. Wilbur Lam, a professor not involved in the study, emphasized that true clinical utility must be demonstrated before it is widely implemented.
People interested in testing the technology can participate in the FaceAge Research Study by uploading selfies; the platform has attracted over 10,000 participants since its launch. Researchers aim to expand their database to improve algorithm accuracy and explore associations between lifestyle factors and perceived age.
The AI tool may help identify frail patients who require additional support before treatments, potentially bridging gaps in care for older cancer patients.